The AI Investor Podcast
Join Eric Bleeker and Austin Smith from 24/7 Wall St as they discuss how artificial intelligence technology is quickly flowing through the global economy - leading to massive changes and opportunities for forward-looking investors. The AI Investor Podcast from 24/7 Wall St. explains, in practical and accessible terms, why AI is such a disruptive and exciting technology and shows investors how they can potentially position their portfolios to benefit from these game-changing shifts.
The AI Investor Podcast
The AI Bear Market Rages On. 3 New AI Stock Buys (Including New Robotics Buys!) Plus, Jensen Is Out at NVIDIA
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
NVIDIA CEO Jensen Huang leaves in a shocking departure. Eric and Austin discuss the bull and bear case for the stock after his exit and what it means for the broader space. Eric then dives deep into recent developments in the AI space, discusses Semi Analysis slashing its estimates for compute spend, talks about how he's thinking about portfolio management with the AI bear market at its nadir, and issues three new recommendations. One that's a repositioning in memory and two that continue establishing a broader base in the robotics theme.
And, you can track Eric's entire portfolio using Yahoo Finance's AlphaSpace tool, one of the most powerful investing tools today that puts a professional research desk in the hands of every user. Get started today >>>>> https://247wallst.com/go/yahoo-alphaspace >>>>>
To view AI 2027: https://ai-2027.com/
0:00 Introduction
0:24 Jensen Huang Out at NVIDIA
8:34 NVIDIA's 'Moats' After Jensen
17:28 The NVIDIA Bull Case Post-Jensen
25:16 The New Reality of China and U.S. in AI
29:50 The Race to RSI and Major Developments in AI
41:51 Semi Analysis Slashes AI Spend Forecasts
45:32 Portfolio Strategy with AI Sentiment at Its Bottom
57:04 ACM Research($ACMR)
1:02:40 Novanta ($NOVT)
1:10:11 Analog Devices ($ADI)
Subscribe on YouTube: https://www.youtube.com/@247wallst.
If you want to work with us at 24/7 Wall St: contact@flywheelpublishing.com
To follow our portfolio: https://247wallst.com/aiportfolio/
To follow Eric on X: https://x.com/bleekertech
Submit your questions: questions@247wallst.com
Join Eric Bleeker and Austin Smith from 24/7 Wall St as they discuss how artificial intelligence technology is quickly flowing through the global economy - leading to massive changes and opportunities for forward-looking investors.
The AI Investor Podcast from 24/7 Wall St. explains, in practical and accessible terms, why AI is such a disruptive and exciting technology and shows investors how they can potentially position their portfolios to benefit from these game-changing shifts.
On today's episode of the AI Investor Podcast, we look at Jensen Wong's affair and what it means next for NVIDIA, AI and the race to RSI, Anthropics, big IPO this week, and most importantly for all of our listeners, we announced three new stock buys in the AI Investor Portfolio. All that and more is next. Eric, uh, where do we begin? Uh there's never been any shortage of intrigue or great headlines in the AI space, but the the palace intrigue, of course, is Jensen Wong getting uh you know honeypotted by potentially a Chinese spy, which then led to this domino effect of you know learning about not just one, not just two, but potentially three affairs. So Jensen Wong is now out as CEO. I'm sure all of our listeners know that. There's a lot of good drama here. Uh it's been covered at length, but I really want to understand, you know, let's just talk very briefly about what happened, just to provide context. And then I really want to understand what this means for NVIDIA, the stock, NVIDIA the company, and if we can get to it, the AI industry overall, because news of Jensen leaving this space just it created a vacuum, and it seems like the bottom fell out of a lot of AI stock. So, Eric, where do we begin?
SPEAKER_01Where do we begin? That's that is truly the question this week. Um, I we we are a few days late. We we weren't able to hit our last Thursday show. I was I was traveling, uh, I was on a cruise with my family, and I just wasn't able to find anywhere that we could really film. And I did want to give a couple days to this Jensen news. Eric,
SPEAKER_00Eric, just be honest. A business executive from Singapore gave you tickets to this cruise. The two of you went on this cruise to discuss pure normal, above-board business matters and um chip chip tariffs. Is that correct? Is that what you meant?
SPEAKER_01I I don't like where this is going. I my wife's already uncomfortable listening to this podcast. But no, uh, you know, we we are we are have a lot to cover today, as you mentioned. We we like to kind of joke like, oh, no, a lot happened last week. I can't think of a crazier week than this. When you look at the amount of black swan events that could have happened in the AI space, the kind of titular figure of AI having an affair with a Chinese spy, essentially bringing down the entire AI complex. I I mean, Austin, what what what could even get weirder from here?
SPEAKER_00Uh well, you know, it what's the expression, right? May you live in interesting times. Yeah, and um, we certainly do, and things certainly could get weirder from here, but I can't project necessarily uh a more this story has all the ingredients, right? You know, you've got an affair, which of course is great for headlines, but Jensen was also in many ways single-handedly pushing the AI industry forward, not just because they're chips, which everybody knows, but because NVIDIA was recycling their balance sheet into other AI companies and just creating this virtuous cycle of, you know, making capital available. So, in many ways, where NVIDIA has been the vacuum that's hoovering up all of the capital from the Mag 7 and the hyperscalers, uh, it then NVIDIA is now taking that capital and recycling it back into the industry. So I want to understand what happens next. And I have I have my own takes, but I also want to be clear, I have gotten this story wrong at every turn that it's been developing. So I want to hear what you think, and then I'll share my opinion.
SPEAKER_01Yeah, and you know, the the headline figure, this this is not good, right? The the AI space, it's it's losing a key figure that was so influential in the creation of AI itself. It's it's someone who has coordinated, as you just alluded to, vast sums of capital, not only for uh, you know, the neo clouds, but also the broader supply chain. And also Jensen Wong's been a visionary willing to work on very long-term bets. We look at CUDA, which is you know basically the framework for making GPUs more general purpose and and the kind of longevity of that bet. I still remember with NVIDIA first kind of recommending them and and kind of buying into the story and in 2009, and and so much of it was kind of this vision that only Jensen Wong was responsible for propelling. So this is someone who has created so much value across his life, and and and he has really been a force in pushing the limits of of what AI itself would be, and uh, you know, suddenly removed uh so we'll we'll have a lot to talk about this, but there's a lot of implications here. There's there's not just NVIDIA, there's a potential new era, and and what does this say about US and China relations going forward? Uh I think it's pretty clear to say that, well, this this felt fairly coordinated. It you know, when when you've got when you've got something that's a coordinated plan that you want someone to leave, it's every single time there is a new news breaking, the moment the oxygen would get sucked out of that, there's there's there's another piece of news, another piece of news. And and that's what kind of the uh various reports about Jensen Wong felt like. Uh, but but what it it the implications are, it's pretty clear we're headed towards a break between a Chinese and an American kind of ecosystem around chips. And just to make it clear, what Jensen Wong did, it it's a perfectly firable offense. Uh clearly he had created enemies. Clearly, as I mentioned, there's this fairly coordinated campaign against them. It's very doubtful he comes back, although right before we start filming, he he did post a tweet saying that he would be open to it, but boards are risk averse, and he created a lot of uh potential liabilities for the board itself. He he clearly wasn't truthful to them, he clearly withheld a lot of information. There's national security implications here. There's there's a lot of sweeping effects. So I think, like you said, let's talk about in the upcoming sections. Let's talk about NVIDIA itself. This is a huge position. It's it's my largest holding. Is it still a buy? How'd you position it? If you own it, what's the future for this company look like? We we need to cover the future of the US and China and AI and and kind of what this event says about where the future is headed, uh, you know, in terms of the rival AI ecosystems between the two uh countries. And lastly, what this says about the future of AI itself, because Jensen was a rabid supporter of uh open weights, of the uh open ecosystem. And uh that that's going to be a place that's going to struggle uh without someone who was kind of a champion of it, like Jensen himself. So I think those are the main areas we really need to hit. And then as you mentioned at the start, we'll also have a few buy recommendations. I'll talk about how I'm looking at stocks at this particular moment, which again, we had we had talked about a broad portfolio recap last episode and said we would move that into some buying this episode. So we'll begin that process today.
SPEAKER_00Yeah, I'm I'm excited to get to that. And the juicy drama here is all about the person, but I'd love to spend more time hearing about the company and the AI industry overall. But with that said, we are in an industry right now where singular people are the deciding factor of whether or not things get done. So I'm thinking about people like Sam Altman, Dario Amadai, uh Jensen, uh Elon Musk. And you have to imagine that Jensen is looking at what happened with Sam at OpenAI as some sort of precedence that some people are so singular that they should, you know, come back as CEO. Now, I'm not going to try and draw an analogy between this situation because it seems like, you know, obviously there's a national security component to what Jensen did, there's also sort of provable um actions, uh, whereas Sam's was largely just uh, you know, a measure of dishonesty. But if the argument is that, you know, this Jensen did things, you know, and hid a lot of information from the board, there is precedent in recent history of CEOs being so singular that this the company votes and the board votes to you know allow them to come back. But I let with that said, I'm sure there's plenty of you know good drama to talk about the people involved here, but that is an interesting dynamic, is like how singular are these dozen or so people leading this industry forward? Let's put that aside. Eric, is NVIDIA still a buy? You said it's your largest holding. You've been talking for weeks correctly about how NVIDIA is cheaper than the market on a forward earnings basis. Yes, this is a cyclical company, but we are also in a cyclical breaking cycle right now, which we have been talking about for years. So what's going on?
SPEAKER_01Yeah, and as you mentioned, when you look at the cheapness, clearly there's a disagreement happening here where NVIDIA is forecasting to 70% growth in its next fiscal year. Well, um, Wall Street is uh pricing it at before Gen Side left, it was still already at 12 or 13 times those forward earnings. So so clearly there is well, Wall Street consensus models. If you look at the sell side, will have shown significant growth in the next year and then moderately, but still relatively strong growth. That's not the way the stock was being priced, right? And and it's because of some of these concerns over what's happening to NVIDIA's moat. So let's just walk step by step through a few key pieces of what makes NVIDIA a strong company. The first one we need to talk about is CUDA. I'd referenced it earlier. Once again, that's that's the framework that NVIDIA created for how you make GPUs general purpose. They built all these libraries for all kinds of fields. It was a massive undertaking and it it really underpinned the beginning of their success. But you look at the same time, companies like Anthropic on their roadshow, and I believe one of the charts that they showed as part of this roadshow process was the fact it used to take them six to 12 months to do a new install of AMD GPUs because AMD has their own equivalent to CUDA, but it's not very good. Well, we are seeing the walls come down in terms of you know, moats for businesses that rely on software extremely quickly. And nothing has come down quicker than CUDA because Anthropic is saying that this process that used to take six to 12 months now takes about three days. So, you know, CUDA was the longstanding. If you had asked me in 2023 or 2021, what is NVIDIA's key competitive advantage, you would have said, well, it's CUDA. It's this software ecosystem that they are just significantly better than their competitors. That was uh, you know, at the beginning of the AI trade, you look at NVIDIA having something like 90% market share in terms of accelerators, you know, these uh chips that do all the processing for these large AI workloads in 2024 and 2025. And what's going to change in the future is it's just become so much easier to get these custom chips working. It's not nearly as complicated as it used to be. AMD's a much better um what would we call it? Uh you know, you'd call it a competition or a thing to replace NVIDIA. All of this has changed rapidly. So we we have the coup demo being, if not gone, extremely rapidly degraded. Next, we have the roadmap from NVIDIA. One thing that Jensen Wong was doing for NVIDIA was he was pushing an extremely aggressive roadmap of annual releases. You know, this is where, again, we need to see how much this is going to be a product of someone like Jensen Wong. Um, with Apple, losing an iconic leader like Steve Jobs. They had a Tim Cook replace him. Cook didn't have the design taste Jobs had. He wasn't able to come up with the next blockbuster products. But what he was able to do was he was able to continue keeping the release cadence on the rails every single year and make sure that supply chain worked. This is a big thing that we're going to need to see for uh Jensen's replacement because right now you can see this already being degraded a little bit. You look at some of the key performance metrics, the total cost of ownership of various systems. We're looking at you know how much token output you get per dollar spent. We see AMD's newest system that's the uh MI455 ahead of NVIDIA on mini metrics. We see uh TPUs moving ahead of them on mini metrics. Why was NVIDIA falling on Monday? Well, part of it's just this continuing negative trend, part of it's uh semi-analysis forecast we'll discuss later, part of it's remote access rules for China, but another part of it is supply chain chatter that Vera Rubin delays um uh are in the making, right? And and the more that NVIDIA has delays uh at this critical moment, the more they're allowing these competitors to basically just be able to chip into their lead. And then Austin, a couple more, and I would say vertical integration. As we've talked about a lot on the show, NVIDIA people think of it as a GPU company, but it's a systems company. Uh almost all their revenue comes from selling these large systems. Uh, they they accomplish this by basically acquiring Mellanox, building out a networking business as well. When they sell these systems, you know, they sell them to NeoClouds and also hyperscalers, but hyperscalers have every incentive to vertically integrate as much as possible. And we see once again in this pre-IPO process for OpenAI and IPO Road Show for Anthropic, them talking up their future competitive advantage being vertically integrating. They want to be able to own the chips, they want to be able to own the hardware layer and make that their key advantage in how they serve chip or tokens, I should say, for a lower cost. So again, NVIDIA they they have this advantage of being vertically integrated, but everyone else is racing to catch up with them. Um, so you know, they're they're their areas where they were kind of buffering this threat was having neo clouds, companies that don't have necessarily the economics to be able to build their own massive chip programs like Google's TPUs or what we're seeing from Amazon with training. Well, with Jensen gone, there's a lot less probability that they're going to keep funding the neo-cloud environment the way that they have. Another hedge against this was open source, which was mostly out of China. You know, I think 70% of total uh token consumption is from Chinese open weight models currently. But we see with this event happening right now, it's it's likely because hocks within both the current administration and also uh the Democratic Party in the US, there's there's some momentum against Wang, which will lead to it's going to lead to probably a situation where we're going to see less um less running of Chinese models across uh uh the American hardware stack. We saw that with the remote roles announced. So, you know, Austin, this is this is just another area that NVIDIA's been walking a tightrope in terms of how they keep this specific advantage going. And a lot of this tightrope was dictated by Jensen, and now he's gone. And the last thing I should note is their their last competitive advantage is just Jensen himself. He started the company, he runs it like a startup. He has 60 direct reports, he is able to execute on unique visions for the future of the AI space. He coordinates the supply chain, you know. He said so many times that every deal that they ever had with Taiwan Semiconductor was a handshake deal. Is the next CEO going to have that same level of trust with partners, right? Um, so I I don't know. I just want to go over that because when we when we go back and we think about each of those parts independently, I think what we see is a lot of the moat that NVIDIA had is heavily dependent on Jen Sin Wong. To a degree, it's it's hard to say any other CEO in America is this intertwined with the advantages of their company. And and again, it comes at this crucial time where due to the degradation of CUDA, NVIDIA's moat is at its narrowest level ever. It has the least ability to not operate at a just phenomenal level. So Austin, uh you know, I'm summing up there, and that that's probably a little scary if you're an NVIDIA investor, but it's also just I I think it's kind of the practical look at just how much losing Jensen will hurt this company.
SPEAKER_00Yeah, I mean, we we need to be eyes wide open here. Jensen is a singular CEO, and that's not to make an excuse for the things that you know he he's done of hiding things from the board or whatever the case may be. It's it's not a commentary on that, just his talent as a CEO and leader, which we've talked about a lot. Um so as an NVIDIA shareholder, right, it makes me nervous. Uh all the things you said. In particular, actually, the the long-term thing that makes me the most nervous is just the cost of compute and efficiency difference between like Blackwell and TPUs, which we have talked about. You know, taken to the limit, um, you just want to be able to produce as many tokens as possible as efficiently as possible. I'm radically oversimplifying, but I think that matters more in the world we're entering into now, where cost of capital is going up. We've talked about that with interest rates. Um, we're seeing that companies like NeoClouds and any other and a lot of other AI companies are running out of financing or having a more not running out, having a more challenging time. So when you're in that environment, you sharpen your pencils a little more and you start to think about things like total cost of ownership, where TPU, I think, you know, ironware, ironwood and trillium are, depending on the measure, like 30 or 40% lower TCO than Blackwell. Blackwell's still far and away the more capable, you know, broad application chip. Um, but let let me let me give you this to react to. And I want to be clear that I have misread this situation from the beginning. I did not think Jensen was going to leave. And um, I also had an early bet that I thought he would be coming back, and that seems to be a fading trade. So I've I've read this situation wrong from the from the onset. But let me put this to you. Let's say new CEOs in charge whenever they're announced. Jensen is singularly out there, you know, investing in all these other AI companies, propping up the AI industry. That money that was going into financing these other companies is now staying in NVIDIA as cash flow. Meanwhile, NVIDIA has a 24 to 36 month runway on deals that are already pre-committed to for things that Jensen has negotiated and they're still commanding a crazy premium. And Jensen, you know, singularly, as we've talked about, um locked up the memory trade in Korea for their high bandwidth memory, at least to the best of their ability. So is there an environment where for the next 24 to 36 months, NVIDIA just pukes cash flow, right? Because now that money is not going out, you know, into NVIDIA and then out to the industry, but it's staying in NVIDIA because the new CEO says, you know, like I need to focus on our operations, I need to make sure we can execute on what we've done. Whereas Jensen, as you said, ran it like a startup and was sort of making those longer shots. Is there an environment where the next 24 months NVIDIA is just like a cash geyser, even more so than it already is, because Jensen is not there?
SPEAKER_01Yeah, well, I you know, you need to be humble, right?
SPEAKER_00That that that's not a long-term call, right? I think long-term you want Jensen and NVIDIA. I'm thinking strictly for the next 24 to 36 months.
SPEAKER_01There is there's a chance we're recording this, and the podcast will be out after NVIDIA's earnings. Um the the fear I would say in the market right now is that they're going to pull their guidance. Uh, especially you just look at what the impacts of of some China revenue and some of the recent policy is, it could lead to them just pulling that. Um it does come back to I think NVIDIA is going to be so much less valued by the broad investor class on how much cash flow they're going to generate this year or next. And and it's almost entirely on what kind of market share they're able to maintain. Um, because again, the the fear is very much you you had a couple things with NVIDIA. You had a company at 90% of basically lab compute across 2024 and 2025. We know that's going down. We we know their lead is going to get chipped into, and that's largely from custom chips. It's partially from from AMD, but but we know that's going to happen. They have to protect both this revenue growth and they have to protect their margin profile. So yeah, they they could do it. You know, they they have some outlets. You you look at their their number one customer is is suddenly SpaceX, as an example, right? Musk has extremely ambitious plans. The market is is forecasting right now that he will not be able to deliver on uh a significant part of those plans, right? Right. You know, he's talking about spending up to 500 billion next year. I don't think anyone expects that in the current environment, right? Not just because of the shortages and inability to get things like power, not just because of the political climate, but also because of the financing situation. That how would you possibly finance it at this moment in time? But a SpaceX has essentially gone all in with NVIDIA. If they wanted to start up a rival program for custom chips right now, it would take a long time to be able to get that up and going, secure the supply, et cetera. So they they do have some powerful potential catalysts, right? But I think it is just going to be so dependent on what kinds of reads are we getting from what kind of market share NVIDIA is going to be able to maintain. The other question from it too is how big is the tailwind at their back and across the entire AI space? Right now, estimates are coming down on how much spend is going to happen the coming years. A headline we'll talk about later in this episode is semi-analysis slashing uh their estimates for future compute, uh, how much is going to be spent on the accelerators that Nvidia and Broadcom and AMD make. Well, a big part of that is they didn't necessarily slash the total number of units by a massive degree. What they slashed is the amount that NVIDIA is going to be able to charge for their upcoming systems beyond Vera Rubin. They they they took a huge just sword to that and slashed NVIDIA's pricing power because their belief is we're going to move from this supply-constrained situation to where the hyperscalers are going to have a lot more leveraging power. So, you know, Austin, if if NVIDIA is able to deliver on their 70% growth this year, and they are in a position that they will still be able to grow the year after that, its valuation is going to look comical today, right? Because they are priced. You know, you look at the broad AI chip trade today, and Goldman Sachs put out that, you know, it is basically baking in right now that we are going to have negative earnings in 2028. That's that's where the overall valuation of this is. If NVIDIA is able to deliver on anything approximating this year's growth and not take a step back in 2028, it's going to be an extremely solidly performing stock from today. So I guess it's just you're asking the scenario analysis. Is there a path that a new leader comes in? He operates NVIDIA tremendously efficiently. They hit their targets on their upcoming releases, they keep their competitiveness versus rival solutions. Yeah, it can definitely work, right? And we're we're gonna need to see who comes in. But but as I said earlier, that that execution has never been more important. Delays right now from NVIDIA, if Blackwell was delayed, hey, that meant that they were going to be pushing some revenue back, but it wasn't existential risk to the company. Right now, being able to stay ahead of the company of the competition is a higher existential risk than NVIDIA scene. And once again, it's because they are losing some of those other competitive advantages.
SPEAKER_00Well, and uh let's not forget one of the biggest levers in the NVIDIA bull story, which the new CEO cannot pull, was uh the ability to sell chips into China. And suffice to say, after this, I think that door is probably closed for a little while. But that was always one of the call options with NVIDIA is well, you know, they'll get an exemption to be able to sell their chips, and there's this, you know, nothing about their forecast includes any China sales. They're not gonna be able to sell to China for the foreseeable future after this. Just not, it's not even gonna be an option for the next CEO.
SPEAKER_01Yeah, I mean, you know, and this is part of why chip stocks are have been going down the past few days, is because we are increasingly looking at a situation. Why, why, where did these leaks come from? Well, it it's almost certainly what you call the China hawks part of the administration because we had two different pathways for who was fighting within Trump's administration and you know, in the seats of power for the US government. And largely what was going on was Trump and Lutnick and some of the other people, they were allowing chips to make their way into China. They were turning a blind eye towards China basically using cloud services, mostly in Malaysia, outside the country. So when you look at an NVIDIA, how much revenue were they getting from China? Well, a lot of people say, well, it was almost nothing. They're not selling their H200s. Well, chips were making it into China. They they did get some sales from legal kind of nerfed down Chinese chips in the past. And then there's also the Chinese companies using cloud services out in Malaysia. So estimates can vary. They go up to 20% of revenues. But as you noted, we have new remote access rules this week that are going to make it so that essentially if a data center located in some of these countries like Malaysia, if they are seeing most of their demand coming from China, they're not going to be able to get future U.S. chips. So this is going to be a relatively large headwind. The other part of this, Austin, is for a while, the strategy that kind of Jensen had favored and and by indirectly kind of Trump did was to allow China to have U.S. chips, but keep them
SPEAKER_00the depend the dependency path that we had talked about.
SPEAKER_01Keep them off the leading edge. And and that way they they would have less incentive to go all in on building their own kind of indigenous supply chain for chips. Well, at this point, we have crossed the Rubicon where we are going to have to see this.
SPEAKER_00But the time to make that strategic decision was four years ago, right? And we had two different approaches, you know, with the Biden administration and the Trump administration. But like you and I have talked about, that sort of back and forth, you know, even if one strategy is better than the other, but we don't know which one, right? But like let's say one strategy is better than the other. That back and forth, that's all it does is just give China, as we have talked about, every single incentive and motivation possible to engineer around it, right? To build their own chips, to get their own lithography machines. That sort of inconsistency just further motivates them to in-house it. So like that horse is out of the barn, right? It's almost like even if you were able to pick the correct strategy today, that in the path to independence is already being paved. So it's it almost doesn't matter.
SPEAKER_01Yes. And it's just it, but this is going to be a threat, right? It means that they are going to push even harder at making their own, you know, domestic chain for memory. They are going to build their own semiconductor equipment chain. And a lot of companies that previously didn't have competition now face competition. You know, an ASML, you face the prospect of EUV years from now. But a lot of the other semiconductor equipment companies, they'll face more competition relatively sooner. So there's just extremely, there's extremely far-reaching consequences from this. It's not just Jensen being removed, it is a signaling the end of kind of one era of this muddled approach to China and the beginning of a new one that's that's going to be significantly different. And that has implications across kind of every single semiconductor company, explains a lot of the it's not just that one Jensen was removed in the past week. It's that there is now a certainty priced into the market of all these semiconductor companies operating in an environment with less to sell to China and more competition down the road.
SPEAKER_00Okay. We've beat a dead horse here. Um so let's transition to what, you know, sort of the future of AI and what the industry is looking like overall right now. So you had mentioned to me that the past couple of weeks are some of the most consequential in AI. There's an essay you wanted to talk about. I also want to get to some of the RSI claims. You know, I was just looking at something before this where Anthropic said some of their internal RSI alarms are starting to ring, you know, so to speak. Not that they've crossed RSI, but they're on the path there and, you know, getting asymptotically close to it. Um let's talk about the AI industry right now. And then I really want to, of course, get to the exciting part, which is the new investments in the portfolio.
SPEAKER_01Yeah, nothing, nothing says excitement like we're going to discuss an essay. Um, this this goes up there with our assigning homework sometimes.
SPEAKER_00We haven't assigned homework in a long time.
SPEAKER_01It has been a while.
SPEAKER_00It has been a while. This is like this is summer vacation for our listeners right now. But you know what? Now that you said it, maybe we need to whip them back into shape. I'll call I'll come with a book recommendation on the next episode. Okay. Can we we'll we'll we'll prime the class.
SPEAKER_01I'll give them a break. I I don't think anyone wants to be doing any homework with the market in these dreary of conditions as we're facing. But the the essay I want to discuss was um AI 2027. We can put the address for that in the show notes. But this was a proposed future from I I don't know what you'd call them, people adjacent to the kind of AI research infrastructure. And it proposed a future for AI where agents make progress in the background throughout early 2027, before a super intelligent agent is created in September 2027, uh, that is able to create a former year's worth of progress every single week. Now, Austin, you can imagine this this came out and it it's not gonna be on you know 2020. You know, this isn't front page news. This was super niche within AI communities, but it basically showed the divide that's going on between one group who looks at a trend line and says, this is when the world could start getting extremely weird, and that's within 2027. And another group who said, Well, you're not factoring in that we're going to hit all of these roadblocks. This is an overly ambitious there, there is no way what is happening in AI 2027 could possibly happen. And I should note where AI 2027 gets deeply weird is at the end of 2027, and that's where essentially you you make two paths where either you're going to pace the progress of AI or you're going to go into a race, which is a little bit analogous to where we are right now, where you have extreme state competition between the US and China. You've got companies that have less oversight. And that's that's not a reality that was pretty. Now, Austin, here's the
SPEAKER_00thing
SPEAKER_01that's weird.
SPEAKER_00Let's just be clear that there is no pacing. It's not gonna happen, right? It is I I I do not believe pacing is gonna happen. I'm gonna plant my flag. People have been calling for it for too long. I don't think I don't think we're gonna see any pacing. I think this race is so far run that you're not gonna get anybody to slow down multinationally or domestically.
SPEAKER_01I will say, when you look at their example, the pacing scenario is still incredibly faster than today. It still gets very weird. Um, and we could we could talk about more, but I think a lot of people when they hear pacing, they think it's maybe a death knell to kind of this broader AI trade. AI can still move significantly faster than today with pacing controls put in. But the thing about AI 2025, we are now, when we look at the trend line laid down the essay, we are running in front of it. We are actually at a faster pace than too. And the background in the past couple weeks, because you talked about this significant news, is Metter, which is an organization that has been embedded in Anthropic is kind of a third party uh reviewing their work. They found that relative to historic levels, AI progress is now moving at a rate of about 2x. And the divergent just started in April of this year, and 80% of this acceleration is downstream from AI. We can we can try and put this up on the screen. I'll I'll try and get this image attached. But what you see in this image is a clear trend line. We have a trend line for years that was creating all the advancements we've seen post-GPT in 2023, 2024, 2025, and we see a clear step change that's happening in 2026. And this is the single strongest piece of evidence yet that we are at the beginning of what could be a massive step change driven by RSI. And the argument essentially goes we are now at a point where you might have hundreds of AI researchers, these are agents, um, and it'll soon be thousands. And and they are less capable than a leading researcher today. They're less capable in knowing what experiments to run. They're sometimes less capable in drawing connections between breakthroughs. They're they're they're less capable in what you'd call general taste. But what Eric,
SPEAKER_00I just want to can we just pause a beat on this because I'm sure our listeners know this, but this is a valuable point, right? RSI is the point where the AI is better at training itself than the current humans are, which results in a takeoff because it can be recursively self-improving itself. It's implied in the name, but I want to be clear that's the the fact that an AI is better at training an AI is a crucial point for that reason because you kind of, in a way, you all you remove one of the final bottlenecks, which is the humans, and then the remaining bottleneck is power.
SPEAKER_01Correct. And this is where these time forecasts are so important because I've said in the show before the the current trend line is that around March of 2027, you will have par where your your average researcher, your median researcher, is about on par with the agents responsible for improving the model. Now, the difference between that would be how many researchers can you have at one spot? You know, how many can you have at OpenAI? You know, maybe, maybe hundreds. Well, you could conceptually have thousands of agents. These agents don't have to sleep, they don't have to take lunch breaks, they would be continuously working on them. And then the current trend line would be that towards the end of 2027, perhaps in you know, August or September, we will have these agents, these autonomous researchers for building the next level of AI. They will be on par with the best human researchers. So all of a sudden, you have all of this compute. You could have tens of thousands of these autonomous agents that are able to research and improve AI itself at a level that is now at or exceeding the best human researchers. And this is exactly what was outlined by this AI 2027 scenario. And it is, again, for someone who read this years ago, it is kind of borderline mind-blowing to see this actually on the trend line that they predicted. So, you know, there's there's two outcomes. Either we are going to see an exponential jump in AI capabilities beginning in 2027, that this 2x rate of improvement that we've seen from these initial reports, this is going to continue and it's going to get even steeper and steeper throughout the year. Or aspects of AI research and improving models, it it just proves to be much more difficult than imagine. You know, there's this is this is what skeptics will talk about, right? That, well, we're we're going to have these autonomous researchers, but they're not going to over be able to overcome some limitations of LLMs themselves. But I mean, Austin, this is this is kind of everything right now. You know, this is when when people look at does the AI trade work, this is the next thing that happens. And we look at this is a tweet. Well, once again, I'll try and get on screen. This is from Pushmeat Coley. I I probably said that wrong, but um, he said, across our projects at Google DeepMind, we are witnessing a genuine phase transition in the pace of scientific discovery. You will see a glimpse of this in our releases in the next few weeks. So, Austin, this is the person who we've talked about, AlphaFold. He was an inventor of AlphaFold. He he was an inventor of GNOME, which is a model discovering millions of new materials. Uh, he's built self-evolving algorithms. This isn't someone who's just going to post engagement bait, right? He's he's not going out there to see if he can get a few hundred tweets, likes on his tweet. He is doing this because there is genuinely something that has flipped within DeepMind that's going to alter the pace of scientific discoveries itself. And and we'll talk a little bit more about Fable 5.2 and our by recommendations. It's a really impressive release. So we are seeing the models that are making out to the public, they are getting better at faster paces, they are getting more capable, and we are seeing more chatter that it's not just anthropic, it's Google, it's open AI all reaching, you know, beginning this process. So the bigger picture right now, when we look at this AI progress, where are we starting to see it? Well, we're filming this right now, and half the world's services are seeing outages, right? I think like Marriott Hotels is out, Apple, uh half their services are out. The NSA's confidential network is taken offline today. Um, that that might be some level of cyber attacks that are announced. That might be the fact that these companies are having to race to patch all their services as more and more exploits are being discovered, and that's leading to more things happening. But this is going to significantly increase the chances we've talked about a lot on the show, that there's going to be some weird stuff that happens and the cybersecurity situation is about to get a lot worse before it gets better. And and second, you know, I would say this all happens at a moment. Once again, we talked about the China Hawks pulling off a coup against Jensen. Um, is this a preview of more consolidation? You know, especially if we have these cyber attacks hit the country, you know, are we effectively down to only a few firms and OpenAI, Anthropic, and Google that that can protect the infrastructure of the country? Is is nationalization a stronger possibility? My my guess is from what we've seen in the past few weeks, there's a much higher probability China is going to concentrate its resources. It's it's going to make some national champions of companies, especially as they are falling behind in in models, especially as they're not able to distill models. Um, so I just wanted to talk about it's it's a weird time because the negativity around AI stocks is worse since we've started doing this show. Without question, it's worse. The evidence that we are about to embark on an era where AI is going to see its strongest growth rates and capability is when I saw that result from Metter and their experience at Anthropic. I mean, Austin, that was like a jaw hitting the floor kind of moment. That was like, okay, there's been a lot of doubts over whether we are in this first inning of RSI, and it is actually yes, it is putting stuff on significantly above trend line, significantly above, and it is creating a scenario that 2027 is the beginning of an explosion in the capabilities of AI and and the abundance of intelligence.
SPEAKER_00So how does that square with some of the new information we saw out of semi-analysis, right? So on one hand, you're saying we're accelerating faster than this meta report, we're actually above trend, weird stuff is going to happen, which is you know very nondescript and ominous. Um, but I but but but I agree. And and and yet, um semi-analysis, who's you know, in many ways been out in front of all of this and showing people just how powerful and how quickly this industry is advancing, looks like they've pulled some of their compute forecasts back. So, uh how are these two things coexisting?
SPEAKER_01Yeah, it is interesting. Uh so semi-analysis, we've obviously talked about them. Research firm, they've been one of the most bullish companies on the the future of AI. Uh, they released new projections, and the headline figures are astounding, taking taking estimates for how much is going to be spent on AI computing, down 40% in some out years. Um, which, you know, we're talking about this at the same time. We're we're talking about some of the biggest advancements in models ever, and it looking like we are on this trajectory to see a wild takeoff at the same time. The single kind of company that's been most closely associated with calling a lot of the the early AI trade is is driving back their expectations. So, a couple things. The headline figures look pretty gnarly on this. Um, we'll try and put on screen, but what their cuts do is it actually puts the company almost exactly in line with consensus. So if you look at this chart, you can see it compared to Morgan Stanley. Um basically, Zemi-Analysis previously was the one research shop that was significantly above everyone else, and now they're just in line with everyone else. So if you see some kind of doom saying, I did look on Twitter and there's a lot of people saying, ha, told you this bubble was popping. It's like, well, it is notable that company was the most bullish, um, is dropping their estimates, but they are just dropping it down to a consensus that still would indicate about 60% growth in 2027 and 50% growth between uh 2028. 2030. So if those growth scenarios played out, the majority of AI stocks are priced significantly below that scenario. So they are not priced to hit where semi-analysis moved their targets to. And second, I allude to this earlier, but much of the cuts when we look at the the specifics. This was only sent out to their institutional clients, but I was able to see what they had sent. It's because they are basically dropping how much NVIDIA is able to sell its chips for. Again, the the generation beyond Vera Rubin, which is Feynman, they have the selling prices dropping from 29 to 44%. They don't have nearly as big of drops for companies like Broadcom and AMD. So, you know, I would just say the headlines on this look extremely scary, but I would just keep in mind this is a company that made its made its name on some extremely bold calls. And sometimes as companies get bigger, uh they don't want to take on the risk of appearing very wrong. So, you know, it's it's kind of cut back to where the herd is, right? There, you're you're not gonna look like an idiot for calling something exactly the way that Morgan Stanley does.
SPEAKER_00Right, which as you had mentioned is still uh merely insane exponential growth as opposed to whatever you know, light speed growth they were forecasting previously. Uh Eric, I know we're light speed. Um I I know we are uh twing on late in the episode here. I want to let let's get to the most important stuff, right? Let's talk about the buys. It's been a while since we've added anything to the portfolio. I know that's what our listeners want to hear, that that's why they're here. What are you doing? What are you adding to the portfolio right now? Because you're and a reminder to our listeners in the past, and we've talked about the deep seek moment many times. There have been other sell-offs and other tremors in the industry. Your best calls have come in those downdrafts, and we are certainly in one of them now. So I'm paying particularly close attention to the buys you're releasing this week.
SPEAKER_01Well, and I think our best buys were when we were able to bet on major transitions away from a lot of exuberance in the market. You know, we haven't done great this year in terms of our recommendations, especially post-March. But, you know, the the reality is we've we've been largely in AI infrastructure, and I just about every AI infrastructure stock is down 30 to 80 percent, you know. And would would I have loved to have done a better job um making some strategic cells? Uh yeah, of course you would have. But um, you know, partially this process is where are we today? And as I allude to in the last episode, I'm thinking a lot less about well, do I want to allocate to memory or optics? I am largely thinking about, okay, what's my allocation to things that are in the broad AI compute bucket? And we we got relatively overbalanced on that because so many of the stocks had run up so much. And how do I want to continue diversifying the portfolio to stocks that have different demand drivers than necessarily this kind of AI compute situation that's become um kind of the defining question of this current moment? So we we did get a comment from someone Jiggy Money, nice, nice uh Twitter name there. And they said, Your followers need you this week, Bleaker Tech. Hope you're feeling better. We desperately need your objective take on the AI trade and how to position ourselves with significant declines. Are we big holders for years or do we cut our losses and move on? And then he took her applied opto electronics, Nvidia, and Marvell.
SPEAKER_00We need to cue the uh Princess Leia meme. You know, help us, Eric Bleaker. You're our only hope. And yet I feel the same. I I I echo that. So, Eric, can you give us your your Jedi mind tricks in the portfolio? Where is where is the force? Where are you using the force to uh make buys right now?
SPEAKER_01Brad, no, no putting me in the uh lay outfit next to job of the hut. No one wants to see that. So um I I think partially it just comes down to a continuum. You look at what happened in 2026, and we had called out at the time that it was feeling too exuberant. You had the bottleneck type stocks, and and that's a situation that relied on this concept of a persistent supply and demand imbalance. As we see some of the air again taken out of the AI trades, specifically because of the financing situation, these companies are going down the most, and we've explained the reason for that. Now, you look at applied optoelectronic versus a Marvell, there's there's a huge gap there. You know, when we recommended applied optoelectronics, the the thesis was this is not a good company, right? We said this is not a good company. They just have a very good temporary situation. And that's one I personally pulled back on in my own portfolio. I wish I would have done a sell for the um for the AI investor portfolio. But, you know, this is probably a situation where something like an applied optoelectronics, it overshot and it's probably overshooting a bit on the way down. But what's its realistic valuation? Probably somewhere, you know, I could be convinced $60 to $80. Probably something like that in a non-super pessimistic scenario. Now, I will say though, I am not saying that because it's trained for like $45. That's something you should go and buy that as a incredibly risky company. What I am saying is I I think a Marvell, this is a company relative to an applied optoelectronics that um is this is a much higher quality company. It's a company with relationships in the right areas, such as custom chips. It has a broad vision of how it's going to be a connectivity leader. It's it's buying lumentum to be able to continue its own vertical integration. I think the the where the market is moving is you did see a lot of what you'd call lower quality companies, ones that had a stake in this kind of bottleneck situation. They ran the highest, they ran the quickest at the beginning of 2026. But where the future is headed is that we're probably going to see more of these higher quality companies that are going to be able to survive, you know, potential potential market cycles. These are the ones that you increasingly just want to make sure are a much higher percentage of your portfolio. So, you know, I mentioned it earlier, but the consensus forecast for accelerators, while the market continues to implode, it is right now still 60% growth in 2027, then 50% more from 2020 to 2030. That's a deceleration, but again, that's still fairly strong growth. Um, and I mentioned earlier that most stocks now are priced for negative earnings growth within two years. We are looking which situation is going to play out. Is it going to be closer to this consensus forecast, or is it going to be that these stocks will increasingly see negative earnings within the next couple of years? And and the question I keep getting from people is the market is so negative right now. What is the next catalyst that that could change the environment? Is that Nvidia's earnings? Is it a strong next revenue figure from an anthropic? And as I talked about earlier, the strongest catalyst is likely what we talked about earlier with um RSI, that AI gets significantly better at increasing rates at in 2027. And that would make the next use cases for the technology a lot more obvious. Whereas right now people are saying, well, there's coding, but what else? So that's the bull case right now, that we are going to see this inflection point in capabilities. It's going to make AI incredibly more obvious in terms of what it can do next. And the market will move from this kind of hand-wringing situation to saying, oh, okay, it's it's really obvious how much better this is getting and lift a lot of kind of the dark clouds overhanging. Now, why could we be wrong, Austin? Well, the acceleration that we believe is going to happen, it might not materialize. Uh, now I would say some of the data we talked about earlier says, I think the likelihood of that is decreasing, but it does exist. We could have financing or kind of geopolitical situations, uh, take a lot of growth out, reduce confidence, create air pockets that make where we're at look like they have a semiconductor cycle. We we could see that AI advances just as much as we thought, but it makes the model significantly more efficient. So the companies that are involved in compute aren't necessarily keeping up, right? Even though AI is increasing, the level of compute needed isn't as much as we thought was. We we could see revenue not keeping pace because companies struggle to operationalize it. Or Austin, we could see another black swan event. I could not have told you two weeks ago that Jensen Wong's removal would tank the AI trade because he had an affair with a spy, right? So things happen.
SPEAKER_00The Netflix documentary is going to be excellent. Let's just let's just agree on that now. The Netflix doc is going to be wonderful.
SPEAKER_01It will finally beat the poop crews. I'm they've been waiting for something to top that. Jensen just handed them on a silver platter. But you know, the most likely Black Swan event I see as I've talked about in the past, a cyber event happens. It it is a significant event. It it cuts off basic services to people across the United States. And and that just further turns the tide in public opinion against AI, even though you need AI to protect against that. Um, it just would turn public tide. So, you know, it would be an indefensible area to to want to pursue if you're a politician. But the key reason I'm talking about all this is once again, I want to uh reduce our total exposure to data center-linked stocks because those positions grew so rapidly, we became extremely overexposed to that. And I believe there's going to be a gradual kind of flight to quality within this space. But I think there are risks surrounding a lot of these companies, even as we get to the situation of RSI, that doesn't necessarily mean that demand for compute will line up with what the projections are for today. So, my goal is if we're buying chips, I want to more reallocate capital. So we're not increasing that amount of the portfolio. I want to add to robotics. I'll explain some more details on this later, why. And I continue to look for new spaces, things like AI use cases, things that are not just chips, that if we imagine forward what this increasing capabilities next year, if we stay on this early trend line are, what are some of the use cases that really benefit? Things like energy, healthcare. I'm also very open to new ideas and continue to look around for them. But these are the areas we are going to focus on. And I do want to keep in mind for everyone out there what we're seeing right now in the past month and a half, it is the single largest defensive rotation in the past 25 years. So there is, of course, we we want to rebalance our portfolio so that we are not out of flux with one risk factor. But at the same time, many of the stocks that people are funneling into today, things like Colgate and a lot of these defensive stocks, if we see the scenario that AI takes off to in in the second half of 2027, if it hits the kind of growth that is at this moment very possible, these are the same stocks that people are probably going to be funneling out of. Because all of a sudden they have an existential risk that the market wasn't deeply thinking about. So Austin, I will say people need to think about there's a risk from AI, there's a risk from where the AI rotation is headed to right now.
SPEAKER_00I I think I forget if it was the last episode or the episode prior, but we talked about the investor tendency to do this. And I think we saw this, I forget it was in 2013 or 2014. There was some some event. I don't know, vaguely in my mind, I think it was the European debt crisis, and people went into Colgate, Coke, Pepsi, all the consumer staples at the times. These companies trade at a premium because they're exceptional companies, but that premium is actually a cost in the long run. And we've talked about that, so I won't belabor the point, but that flight to safety um will hurt on the way up dramatically. Uh Eric, let's get on to the next investment, um, which is boring, but boring is beautiful. So what's going on? Where are you putting where's your second buy?
SPEAKER_01Well, I guess this would be the first one. It's ACM research. Um, so we're we're gonna do first ACM research and then we're gonna look at some robotics buys. Um, but ACM, I I talked about in the last episode as a company I've wondered if I've if I'm kind of missing something or if there's a genuine opportunity that the market's handing me. Um what they deal with is it's boring areas of semiconductor equipment, equipment that cleans, coats, heats, treats, adds uh electrical connections to wafers. Um but what they have is it's it's an American company, but they have a subsidiary, ACM Shanghai, and they they have an incredible toehold into the Chinese market, selling not only this broad kind of equipment, but also some key growth areas such as advanced packaging that China specifically is going to need. Um, because advanced packaging is a way China currently doesn't have access to the most cutting-edge, what you call nodes. They can't produce uh two nanometer chips and three nanometer chips and five nanometer chips, things enabled by EUV. So, a way around this is often that you're going to use advanced packaging to make your chips more capable. And ACM is a company that has a lot of products in this category beyond their other products, which continue growing as well. So, we talked about this week what are the downstream implications of this kind of Jensen Wang saga, and it's it's this decoupling from the US and China and them building their own kind of indigenous supply across all semiconductors. And this is where ACM research is going to benefit, and they're trading down close to $40 per share. They're probably going to earn around $3 per share in the next year, and they they have upside once again if this kind of domestic buildout makes a big push, which I believe the odds have materially increased in a significant way in the past few weeks. So the odds of a massive tailwind at their back have gone up, but their stock price has gone down. So, you know, there's there's just a lot of negativity, Austin, around Chinese AI companies. And there's a reason for that. Again, they're they're falling behind in the race. They're they're leading model companies, haven't been able to keep up as competitively. A big reason for that is companies like Anthropic cutting off uh them being able to distill their models. But the bottom line is AI is going to remain a strategic priority for China. We know that the government sees this as an existential race that they need to stay involved with. And their only solution moving forward is going to be they need to create a domestic supply chain for semiconductors. We I don't need to get into all the comparisons, but we've seen China do this in electric cars and solar panels, and we could go on for everything. And this is going to be the next area they do this. So, what I want to do is I want to sell $10,000 of our Micron stake, which is highly appreciated. I want to rotate that into more ACM research because again, I see this as a situation that likely could create downsides for Micron over the coming years, and it creates upsides for ACM research. So we'll just rotate into that. If there's anyone out there who owns Micron is wondering if if if they should follow along, well, I will say we are we are close to the end of the tax year. So if you're Micron, if you bought it when we recommended it, it's still up 700%, highly appreciated. You can get into January. I don't think, I don't think this is something you need to do today or tomorrow. Uh, you know, um negativity is still kind of the the ruling factor of the day. But I I do like being able to reallocate amongst some of our names that are kind of levered to AI data centers.
SPEAKER_00Um so with ACMR Great Rec, uh, does this also have Lumentum potential in that it's only a only quote unquote a $3 billion company? I mean, a relatively easy takeout target at this size as well. Do you think that's there's that sort of potential for investors? I know that's not why you're buying it. You should never buy a position just on the on the premise of a acquisition, but uh having seen what that did, you know, to lock in the gains on your Lumentum position, you know, is there a similar potential here given the size?
SPEAKER_01Yeah, it's complicated because it is a Chinese company. Um, well, it's an American company, but with a Chinese subsidiary. Honestly, my my one of my biggest risk factors would be if if the competition continues heating up and relations cool between the US and China, uh, you know, there could be situations where they increasingly nationalize parts of the supply chain, and and that could be bad for investors. That, you know, maybe it's a 5% probability, but it is some probability that exists out there. So I think of them less as an acquisition story and just more. This is gonna be a company that has a very clear path to um earnings more than doubling in the the next two or three years and is priced at you know 12 times next year's earnings. So if if they continue to deliver and this tailwind gets very clear, there's massive negativity around the stock today, but I believe over time that that value will be bared out.
SPEAKER_00Wonderful. Uh okay. Well, yeah, two of the greatest early investments, if people were pressured enough to pick them up, are Taiwan semiconductor and NVIDIA, even given you know the fall, the pullback that we've seen in recent weeks with the honey pot scandal. But uh let's look to the next mega industry that we believe uh could be enormous, which is robotics. So, what are the NVIDIA and the Taiwan semiconductor equivalent in this industry? You know, if if you wanted to be early and you were early and you bought up uh NVIDIA and Taiwan semi, you were happy. Let's see if we can make an analogous move in the robotics industry.
SPEAKER_01Yeah, and the it's funny. I I allude to it earlier. Fable 5.2 came out. This is a model from Anthropic. Normally, this would have been one of our leading things that we talked about. It's a very material event. We haven't even covered it till now because there's just so much other stuff happening, but we'll we'll try and put the benchmarks on the screen for YouTube. Um, just some really stunning advancements. Uh, the remote labor index, this is how good AI is at general knowledge work. It shot up from uh 17.9% in Fable 5.1 to 41.3%, massive jumps in agentic coding, agentic scientific research, the kinds of things we're talking about with uh RSI capability, and a massive jump in other areas such as robotics on the benchmark for robotic scoring, 64.9% versus 27.6% for the prior model, Fable 1. So if you've studied AI for a long time, you've likely heard of the bitter lesson, and that's AI systems improving with computation. So we've got LLMs, they will outperform areas with painstaking expertise. So the question is, when would this come for robotics? Because right now, most of robotics is specialized around foundational models narrowly applied to robotics. Is there a time that these frontier LLMs will just get so good from all their computation that they will overpower much of the specialization across the field? And that's that's a space where a lot of the people who are researchers in the space would say, no, this isn't going to happen. But sometimes, you know, it's it's it's hard when when you are so focused on one narrow path, it's it's hard to see kind of the storm that's brewing on the horizon. And Fable 2 5.2 feels like a moment similar to again what we saw with progressions of LLMs and what that would mean for memory in 2025. You you could see what was coming, and you need to squint a little bit, but if it came, how these LLMs were changing and what that meant for memory, it was going to be something very big. It was going to quickly become apparent to the world how much kind of the these changes would would shift how much memory would be needed. And and I think next year, if RSI leads to the substantial gains we talked about, if these models get on a new capability trajectory throughout 2027, the most applicable market for LLMs becomes robotics because their total addressable market is just so massive. So what what robotic stocks at this point becomes is number one, a bet on this acceleration and that robotics will be the most logical market for these model companies like Anthropic and OpenAI to really focus deeply on. And second, if RSI leads to a scenario where these models are significantly more efficient, which again could cut into demand for AI infrastructure, well, many of the companies in the robotic supply chain, they've got significantly different demand drivers, right? So we're not we're not picking up, if we go to robotics, another company that if total demand for you know NVIDIA chips goes down, they're going to fall with them. We're picking up companies that often have significantly different demand drivers. And this is going to mean that, well, we can reduce overall portfolio volatility, which at a moment like today would be fantastic. And we can also position towards a trend that we believe could see exponential progress in 2027. Again, much the way at the beginning of 2025 we were able to call something like memory by just reading where the logical outcome of some of the breakthroughs happening in the space were. So to date, we've got five robox stocks: Regal, Rexnord, Timkin, those are diversified industrials that would have heavy exposure if robotics really took off a significant way. We have Allegro, which has a DAS and our catalyst, but also Roblox exposure. We have Ouster, which is a sensing play, and we have ST Micro, which is a very diversified play. And what I wanted to do today was try and isolate some cornerstones of robotics. You know, you'd mentioned in the tee up to this question, the Taiwan semi, NVIDIA Broadcom. When we started the AI investor portfolio, we we went for each of those companies very early because they could be cornerstones. They were just big companies that if the general idea of AI computing took off, they were very likely to take off with it. And we want to find some similar for robotics. So uh stock one, we'll do Novanta uh or Novanta, I guess I should say. Uh, this is this is a stock that jumped on robotics enthusiasm.
SPEAKER_00This is ticker symbol and N O V T.
SPEAKER_01Thank you for that. And we'll also try and get some stats about on the screen as well for those watching on YouTube, or I should say, with video on Spotify and Apple Podcasts as well. But like many stocks, all the gains are gone, right? There's tremendous enthusiasm about robotics. Um, but that has now been washed out, and and that's that's what we're looking to today. And that's the opportunity because the stocks back to training are a five-year low. Uh, what it does, it makes force and uh torque sensors, robotic tool changers, frameless motors, servo drives. I mean, the bottom line, there's not a lot to overthink here. Does it
SPEAKER_00get any more exciting than that? Does it get any more exciting than that? You know, the robotics
SPEAKER_01is so exciting, and then you break down what it does, and it's like, oh, it just fits a concentrated physical component exposure. Um, and and that's what we're looking for, right? It's end markets are 60-40 between medical and advanced industrial. It's got facilities in North Carolina, it it it's it's not completely decoupled from China, it's got uh it does have uh physical presence there in the supply chain. But Austin, this is a company that the downside, if robotics doesn't really see this takeoff that I'm predicting in 2027, where it's going to become readily apparent that these frontier models from anthropic and open AI, they're going to apply the bitter lesson to robotics itself. And there's gonna be a lot of excitement about how robotics can take off. Well, they're gonna have the upside from that. But again, this is a company trading near its five-year lows. The biggest risk for this and many of these other diversified robotics companies is simply if there's an economic downturn, there's a recession, they're gonna drop alongside other companies. Um, but again, $10,000 into this stock. I just want to keep diversifying into robotics. So that's our second buy today and our first robotics buy.
SPEAKER_00All right, we are on the home stretch here. And um tell me about buy number three. I'm not even gonna tee it up, just jump in. I want the third one.
SPEAKER_01Uh, that's analog devices, uh ticker symbol, ADI. This is not a home run swing at all. Um, as I noted at the beginning, we are we are just trying to form a base. We are trying to form some more diversification into other areas. Uh analog devices, it's a company with the most complete portfolio in sensing, precision motion, power management. Their humanoid content, it's quoted at several thousand dollars per robot. And that's about 10x the content in current advanced robotics. So this is a company with relatively light exposure to DASRs. They have some exposure. It's about 15% of their total revenue. But what they're really working off is end markets like industrials and and automotives. And if we had this scenario of extreme robotics progress, well, I think that would likely begin accelerating their industrials' content as as well. So, again, a company that's just very aligned with if robotics took off, this would be a company like when data centers took off, people first looked to NVIDIA. It was just the obvious candidate. I believe analog devices becomes the obvious candidate for robotics. Now, what I don't love, it's a little expensive. It's trained for 23x forward earnings. You know, if you believe we are nearing the peak of a cycle, that's that's a heavy rate to pay for this company. It's it's heavily linked to the broader economy. So, once again, if we do see a recession, this is this is a company that that could fall back. Now, what I do love, again, that clean connection to the advancements of robotics. Also, winners win, right? And and stocks that have won for a long time generally keep on winning. That's that's a credo that we've had about the stock market. And this is a company that's had consistent gains across the past 20 years in several market cycles. Um, you know, that's going to be embedded within the organization. People, people, this isn't some new startup. You know, this is a company that has seen it all and it's extremely well managed. So we'll do a $15,000 investment in this company, analog devices. We this is one we could have upsized even more. Um, I'm I'm keeping it to that level just because it's it's one that if we keep seeing a sell-off in the market, it's one that we could add to. Um, but but for now I'm I'm comfortable keeping it at that level. So just to recap, we've got ACM research. We are going to sell a small portion of our micron to continue diversifying into them. That's a company with a tailwind from you know these recent events and and what says about China-U.S. relations. We also have um uh Novanta and we have um I'm blanking now, analog devices. So there we go. Three new investments today. And uh as I noted, you know, the goal is to continue diversifying the portfolio, and we will we will seek to do that in future episodes. This isn't a final buy at all. This is kind of just the beginning of continuing to rebalance the portfolio.
SPEAKER_00Um, wonderful. Well, Eric, thank you so much for that. I know a marathon episode, it takes an immense amount of time and effort on your end to make these picks. I know people people might just want the ticker symbol, but you need to understand the many dozens of hours that goes into recommending each of these. So, not just three buys, but also one sell. So, sort of some bonus portfolio activity out there. Um, Eric, thank you so much uh for your time. Thank you for the recommendations, thank you for the portfolio updates. I know our listeners really appreciate it, particularly in this time of volatility. Um, I guess that's it for this week. Dear listeners, thank you for tuning in. And Eric, I will see you next time.